From dean at aspectprogramming.com  Mon Feb 25 22:34:24 2008
From: dean at aspectprogramming.com (Dean Wampler)
Date: Mon, 25 Feb 2008 21:34:24 -0600
Subject: [Aquarium-users] V0.4.0 Released
Message-ID: <2F92DC82-A552-4289-9240-01A46E7390C9@aspectprogramming.com>

Greeting,

I just released V0.4.0, which adds support for JRuby. Not only do the  
regular "pure Ruby" Aquarium specs run reliably under JRuby (V1.1RC2),  
but you can now advise Java types with Aquarium!

There are some *important* limitations, though. Cartographers of old  
would mark dangerous or unknown sections of their maps with "hic sunt  
dracones" (here be dragons), a reference to the practice of adorning  
ornate maps with serpents around the edges.

This is true of Aqurium + Java types in JRuby, too, at least  
currently. See the details here: http://aquarium.rubyforge.org/jruby.html

Yours,
dean

Dean Wampler, Ph.D.
dean at objectmentor.com
http://www.objectmentor.com
See also:
http://www.aspectprogramming.com  AOP advocacy site
http://aquarium.rubyforge.org     AOP for Ruby
http://www.contract4j.org         Design by Contract for Java5



-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://rubyforge.org/pipermail/aquarium-users/attachments/20080225/8467b20b/attachment.html 

From mvyver at gmail.com  Thu Feb 28 00:08:52 2008
From: mvyver at gmail.com (Mark Van De Vyver)
Date: Thu, 28 Feb 2008 16:08:52 +1100
Subject: [Aquarium-users] Aquarium: around classes and descendants
Message-ID: <389c43e40802272108q79ab84b6h731cbe38723a4edb@mail.gmail.com>

Hi,
Can anyone shed light on what I'm doing wrong in the attached minimal example.

What I see is:

"Initializing MyModule::OtherModule::MyDescendant  ancestor"
"Initializing MyModule::OtherModule::MyDescendant"
W, [2008-02-28T16:02:04.641217 #11518]  WARN -- : Warning: No join
points were matched. The options specified were [:around,
{:classes_and_descendents=>MyModule::OtherModule::Abstract,
:in_objects=>MyModule::OtherModule,
:default_objects=>MyModule::MyClass, :calls_to=>:initialize}].

What I expected to se was somthing like:

Entering....
"Initializing MyModule::OtherModule::MyDescendant  ancestor"
Leaving...
We really should see this: ....
Entering....
"Initializing MyModule::OtherModule::MyDescendant"
Leaving...
We really should see this:....


A follow-up question would be:
Is it possible to invoke only around the descendants and not the
Ancestor/Abstract's initialize?

Appreciate any tips.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: scratch.rb
Type: application/x-ruby
Size: 1127 bytes
Desc: not available
Url : http://rubyforge.org/pipermail/aquarium-users/attachments/20080228/fe05f13a/attachment-0001.bin 

From mvyver at gmail.com  Thu Feb 28 01:28:56 2008
From: mvyver at gmail.com (Mark Van De Vyver)
Date: Thu, 28 Feb 2008 17:28:56 +1100
Subject: [Aquarium-users] Aquarium: around classes and descendants
In-Reply-To: <389c43e40802272108q79ab84b6h731cbe38723a4edb@mail.gmail.com>
References: <389c43e40802272108q79ab84b6h731cbe38723a4edb@mail.gmail.com>
Message-ID: <389c43e40802272228k23c317e5g22ed7c7afe85d91@mail.gmail.com>

Hi,
I've made some progress to get to where I wanted... see the attached.

Essentially I wanted to be able to pass an instance of the 'observed'
class to instance methods of the 'observing' class.
The attached shows how I did this.

One note is that if we replace self.new with self.initialize in the
'after' stanza then an exception is raised.
Is this a bug?

All that remains is to see if I can place the aspect around
descendants of a class and not the parent class itself.... is this
possible?

Thanks
Mark




On Thu, Feb 28, 2008 at 4:08 PM, Mark Van De Vyver <mvyver at gmail.com> wrote:
> Hi,
>  Can anyone shed light on what I'm doing wrong in the attached minimal example.
>
>  What I see is:
>
>  "Initializing MyModule::OtherModule::MyDescendant  ancestor"
>  "Initializing MyModule::OtherModule::MyDescendant"
>  W, [2008-02-28T16:02:04.641217 #11518]  WARN -- : Warning: No join
>  points were matched. The options specified were [:around,
>  {:classes_and_descendents=>MyModule::OtherModule::Abstract,
>  :in_objects=>MyModule::OtherModule,
>  :default_objects=>MyModule::MyClass, :calls_to=>:initialize}].
>
>  What I expected to se was somthing like:
>
>  Entering....
>  "Initializing MyModule::OtherModule::MyDescendant  ancestor"
>  Leaving...
>  We really should see this: ....
>  Entering....
>  "Initializing MyModule::OtherModule::MyDescendant"
>  Leaving...
>  We really should see this:....
>
>
>  A follow-up question would be:
>  Is it possible to invoke only around the descendants and not the
>  Ancestor/Abstract's initialize?
>
>  Appreciate any tips.
>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: scratch.rb
Type: application/x-ruby
Size: 1380 bytes
Desc: not available
Url : http://rubyforge.org/pipermail/aquarium-users/attachments/20080228/9bdd5412/attachment.bin 

From mvyver at gmail.com  Thu Feb 28 01:45:26 2008
From: mvyver at gmail.com (Mark Van De Vyver)
Date: Thu, 28 Feb 2008 17:45:26 +1100
Subject: [Aquarium-users] Aquarium: around classes and descendants
In-Reply-To: <389c43e40802272228k23c317e5g22ed7c7afe85d91@mail.gmail.com>
References: <389c43e40802272108q79ab84b6h731cbe38723a4edb@mail.gmail.com>
	<389c43e40802272228k23c317e5g22ed7c7afe85d91@mail.gmail.com>
Message-ID: <389c43e40802272245g696de7a6le861dff7b733f04e@mail.gmail.com>

On Thu, Feb 28, 2008 at 5:28 PM, Mark Van De Vyver <mvyver at gmail.com> wrote:
> Hi,
>  I've made some progress to get to where I wanted... see the attached.

Spoke too soon.... to get the behavior I'd like it is neccessary to
insert a 'non-standard' self at the end of the observed class'
initialize method.
Is this intended behavior?

Another revised version is attached that illustrates a use-case that
might not be too unusual:
 - initialize an instance of the 'observing' class before initializing
the 'observed' class.
 - pass the observed instance to the observer instance.
 - invoke some observer instance methods.

HTH?
Mark


>  Essentially I wanted to be able to pass an instance of the 'observed'
>  class to instance methods of the 'observing' class.
>  The attached shows how I did this.
>
>  One note is that if we replace self.new with self.initialize in the
>  'after' stanza then an exception is raised.
>  Is this a bug?
>
>  All that remains is to see if I can place the aspect around
>  descendants of a class and not the parent class itself.... is this
>  possible?
>
>  Thanks
>  Mark
>
>
>
>
>
>
>  On Thu, Feb 28, 2008 at 4:08 PM, Mark Van De Vyver <mvyver at gmail.com> wrote:
>  > Hi,
>  >  Can anyone shed light on what I'm doing wrong in the attached minimal example.
>  >
>  >  What I see is:
>  >
>  >  "Initializing MyModule::OtherModule::MyDescendant  ancestor"
>  >  "Initializing MyModule::OtherModule::MyDescendant"
>  >  W, [2008-02-28T16:02:04.641217 #11518]  WARN -- : Warning: No join
>  >  points were matched. The options specified were [:around,
>  >  {:classes_and_descendents=>MyModule::OtherModule::Abstract,
>  >  :in_objects=>MyModule::OtherModule,
>  >  :default_objects=>MyModule::MyClass, :calls_to=>:initialize}].
>  >
>  >  What I expected to se was somthing like:
>  >
>  >  Entering....
>  >  "Initializing MyModule::OtherModule::MyDescendant  ancestor"
>  >  Leaving...
>  >  We really should see this: ....
>  >  Entering....
>  >  "Initializing MyModule::OtherModule::MyDescendant"
>  >  Leaving...
>  >  We really should see this:....
>  >
>  >
>  >  A follow-up question would be:
>  >  Is it possible to invoke only around the descendants and not the
>  >  Ancestor/Abstract's initialize?
>  >
>  >  Appreciate any tips.
>  >
>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: scratch.rb
Type: application/x-ruby
Size: 1441 bytes
Desc: not available
Url : http://rubyforge.org/pipermail/aquarium-users/attachments/20080228/5524df13/attachment.bin 

From dean at aspectprogramming.com  Thu Feb 28 10:44:16 2008
From: dean at aspectprogramming.com (Dean Wampler)
Date: Thu, 28 Feb 2008 09:44:16 -0600
Subject: [Aquarium-users] Aquarium: around classes and descendants
In-Reply-To: <389c43e40802272245g696de7a6le861dff7b733f04e@mail.gmail.com>
References: <389c43e40802272108q79ab84b6h731cbe38723a4edb@mail.gmail.com>
	<389c43e40802272228k23c317e5g22ed7c7afe85d91@mail.gmail.com>
	<389c43e40802272245g696de7a6le861dff7b733f04e@mail.gmail.com>
Message-ID: <8969AF02-AC3A-4F0A-A45E-598A24433F8C@aspectprogramming.com>

Hi, Mark,

I found a bug looking at this example :( and two cases where Aquarium  
should provide warnings. I've attached my version and I'll post the  
appropriate bug and feature requests to trac.

If you use :exclude_classes => ::MyModule::OtherModule::Abstract, it  
will bypass this parent module and only advise the descendents, as you  
desire.

Your initialize needed to return "self"; otherwise it would return the  
result of the print statement, which is nil. The better way to handle  
this case is to ignore the return value of initialize, as returned by   
"proceed", and instead simply assign "object" (one of the advice  
block's parameters) to your @observed_instance. It is standard Ruby  
behavior for "new" to ignore the return value of initialize, so this  
is a more "robust" approach.

One thing to be careful about is the return value of around advice.  
Normally, you would capture the value returned by "proceed" and return  
that at the end of the advice. So, you could add  
"mc.observed_instance" or "object" at the end of the advice. However,  
for reasons I just mentioned, this isn't necessary when advising  
initialize methods.

Here's the most curious thing about your example; the ":in_objects"  
you passed masked a bug, but didn't actually match any join points as  
you intended. First, I need to add a warning that you should only use  
this flag with instances of a class and never with a type. Yes, types  
are Objects, but using this flag this way could lead to confusing  
behavior with the metamodel. I need to add a warning against doing this.

In your case, this flag actually matched no join points, because there  
is no initialize method on types themselves, just instances of them. I  
need to add a warning when any explicitly-specified types or objects  
don't actually match any join points. (You do get warned if nothing at  
all matches.) Note that I added print statements to show the  
"specification" (computed from the inputs) and what join points got  
matched.

However, this flag did work around a bug!  There is a provision in  
Aquarium that if you embed an aspect in a type, as here, it will use  
as a "default object" instances of that type. This is so you don't  
have to explicitly name the type when writing advise like this:

class Foo
	...
         # The following 2 after statements are equivalent:
	# after :calls_to => :initialize, :in_type => Foo do; ...; end
	after :calls_to => :initialize do; ...; end
         ...
end

Unfortunately, it appears that aquarium is still trying to advise this  
"hidden" default object (you can see it actually in the  
specification). I noticed the bug when I added the print join points  
statement AFTER commenting out your ":in_objects", which I thought was  
doing nothing. In fact, that option suppressed the default object  
feature. When I removed your flag, Aquarium advised this object, even  
though other TYPES where specified, and that actually leads to an  
infinite recursion when the advice runs!

One workaround is to add MyClass to the list of excluded types, as  
shown (until I fix this bug). Another workaround is to move the aspect  
creation outside of MyClass.

Sorry for the long-winded reply, but you've showed me ways to make  
Aquarium better!

dean



On Feb 28, 2008, at 12:45 AM, Mark Van De Vyver wrote:

> On Thu, Feb 28, 2008 at 5:28 PM, Mark Van De Vyver  
> <mvyver at gmail.com> wrote:
>> Hi,
>> I've made some progress to get to where I wanted... see the attached.
>
> Spoke too soon.... to get the behavior I'd like it is neccessary to
> insert a 'non-standard' self at the end of the observed class'
> initialize method.
> Is this intended behavior?
>
> Another revised version is attached that illustrates a use-case that
> might not be too unusual:
> - initialize an instance of the 'observing' class before initializing
> the 'observed' class.
> - pass the observed instance to the observer instance.
> - invoke some observer instance methods.
>
> HTH?
> Mark
>
>
>> Essentially I wanted to be able to pass an instance of the 'observed'
>> class to instance methods of the 'observing' class.
>> The attached shows how I did this.
>>
>> One note is that if we replace self.new with self.initialize in the
>> 'after' stanza then an exception is raised.
>> Is this a bug?
>>
>> All that remains is to see if I can place the aspect around
>> descendants of a class and not the parent class itself.... is this
>> possible?
>>
>> Thanks
>> Mark
>>
>>
>>
>>
>>
>>
>> On Thu, Feb 28, 2008 at 4:08 PM, Mark Van De Vyver  
>> <mvyver at gmail.com> wrote:
>>> Hi,
>>> Can anyone shed light on what I'm doing wrong in the attached  
>>> minimal example.
>>>
>>> What I see is:
>>>
>>> "Initializing MyModule::OtherModule::MyDescendant  ancestor"
>>> "Initializing MyModule::OtherModule::MyDescendant"
>>> W, [2008-02-28T16:02:04.641217 #11518]  WARN -- : Warning: No join
>>> points were matched. The options specified were [:around,
>>> {:classes_and_descendents=>MyModule::OtherModule::Abstract,
>>> :in_objects=>MyModule::OtherModule,
>>> :default_objects=>MyModule::MyClass, :calls_to=>:initialize}].
>>>
>>> What I expected to se was somthing like:
>>>
>>> Entering....
>>> "Initializing MyModule::OtherModule::MyDescendant  ancestor"
>>> Leaving...
>>> We really should see this: ....
>>> Entering....
>>> "Initializing MyModule::OtherModule::MyDescendant"
>>> Leaving...
>>> We really should see this:....
>>>
>>>
>>> A follow-up question would be:
>>> Is it possible to invoke only around the descendants and not the
>>> Ancestor/Abstract's initialize?
>>>
>>> Appreciate any tips.
>>>
>>
> <scratch.rb>_______________________________________________
> Aquarium-users mailing list
> Aquarium-users at rubyforge.org
> http://rubyforge.org/mailman/listinfo/aquarium-users

Dean Wampler, Ph.D.
dean at objectmentor.com
http://www.objectmentor.com
See also:
http://www.aspectprogramming.com  AOP advocacy site
http://aquarium.rubyforge.org     AOP for Ruby
http://www.contract4j.org         Design by Contract for Java5



-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://rubyforge.org/pipermail/aquarium-users/attachments/20080228/ca988202/attachment-0002.html 
-------------- next part --------------
A non-text attachment was scrubbed...
Name: scratch.rb
Type: text/x-ruby-script
Size: 2914 bytes
Desc: not available
Url : http://rubyforge.org/pipermail/aquarium-users/attachments/20080228/ca988202/attachment-0001.bin 
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://rubyforge.org/pipermail/aquarium-users/attachments/20080228/ca988202/attachment-0003.html